PLEASE NOTE - ALL MY CELEBRITY LORAS WILL BE DELETED (BY MYSELF) BY THE END OF JUNE 2024.
NOTE: I am so exhausted trying to create women's apparel that is modest enough to not get flagged by civit's filters (1cm of skin below the middle of the throat is enough to trigger it sometimes, and at other times it seems to be turned off entirely), that I have decided henceforth to only post face images.
Use a LoRA strength between 0.1 and 0.15. Anything more than that and you will get absolute garbage (see 'the three-times trained method' below).
You may need to use negative prompts about male characteristics such as '1boy', and 'man', and 'ripped', etc (see 'This model is an interesting failure' below).
The addition of the word 'female' and traditionally female-related positive prompts can help to stop Stable Diffusion trying to interpret Jamie Lee Curtis as a man (see 'This model is an interesting failure' below).
If you want to reproduce the classic look of Jamie Lee Curtis, i.e., the short spiky hair of Trading Places and other mid-80s movies that she did, you will have to emphasise this, perhaps multiple times in the prompts, because the use of female emphasis in prompts is likely to cause Stable Diffusion to make Curtis's hair longer, and in general to conform her to its own idea of female beauty.
This model is very likely to produce ????? renderings unless counter-prompted.
Trained three times, at 512x512, 512x768, and 768x512 (identical data, different crops), and then merged with all three models at 100% strength, in Kohya (which is why LoRA strength has to be at 0.1).
1. I diritti dei modelli ripubblicati appartengono ai creatori originali.
2. I creatori originali che desiderano reclamare il proprio modello devono contattare il personale di SeaArt AI tramite canali ufficiali. Clicca per rivendicare
